People often forget about Facebook Groups. They’re low profile, and fairly simple. But they have over 1 billion users. Anecdotally it seems they are influential over everything from private business discussions to the hottest parties in Los Angeles.
This makes Bots for Facebook groups interesting. It will be possible to engage in a Group without being allowed in. For instance BT Sport might create a Bot with sports results for private football groups. Or a software company might create a news bot for a specialist set of customers.
On that note, it turns out the most effective ad of the Trump campaign used an old video of Michelle Obama attacking Hillary Clinton. It’s routine for political campaigns to follow their opponents everywhere, hoping to catch them off-guard. I wonder how many corporates have considered this?
None of us want to get left behind. But technology is particularly prone to damaging hype. Some of it is completely ridiculous. So I’ve written a brief guide to spotting hype. In short, keep your eye out for things that are impossible, bad value or without enough reach.
This week a series of brands decided that the risk of appearing next to ISIS on YouTube was too risky. So they pulled their advertising.
The debate about Cambridge Analytica, Trump’s election and data continues, here and in many other places. All the evidence I can see is that they did smart testing, but nothing particularly innovative. I’m writing up a presentation on it this weekend – let me know if you want a sneak preview of a draft.
The brilliant Venkatash Rao has launched a consulting business with a difference. You subscribe and get unlimited help, by Slack. The big question for me is whether this scales – but I could see some interesting alternative models here from those offered by traditional consulting firms.
There are really two ways to fix a problem like this.
Option 1: Whitelist appropriate YouTube channels
The first way is to whitelist YouTube channels that advertising appears against. For instance you get the most popular 100,000 YouTube channels and put them through a verification process that weeds out inappropriate content. Human tagging, combined with machine learning (in the literal sense), would make this expensive but not impossible.
However the, say, 1.9m channels which would not get verified are 95% of the channels, but likely only, say, 20% of the views. So YouTube would need to give up that ad inventory. Or it could sell it to less squeamish advertisers at a lower rate.
Option 2: Find inappropriate YouTube videos
The second option is messier but less costly for YouTube. It relies on the fact that people who put up ISIS videos want them to be seen. So YouTube can put human review in for any video that:
Isn’t from a known safe channel or user.
Has any tag or keyword that is risky (e.g. ‘Iraq’).
Has more than, say, 400 views (And so is likely to have any meaningful impact)
Is gaining views quickly.
Add these together and the number of videos to review drops dramatically. Get your reviewing mechanism to learn from human judgement, and you should, fairly quickly, be able to handle the problem.
YouTube/Google have some of the world’s best machine learning experts, huge budgets and good reasons to move quickly. They can fix this.
Technology companies constantly hype their own processes and products. Often this hype is technically truthful, but misleading. But hype takes on a life of its own, misunderstood or exaggerated. Eventually hype leads to absurd claims. These cause real damage.
To protect yourself look at three laws of marketing hype.
Law one: Impossibility – The technology doesn’t exist or is impractical
Impossible claims are at the centre of technology hype. Or the hype focuses on things that might be technically possible, but impractical.
Mass scraping of data is usually impossible
First consider a Guardian claim about Cambridge Analytica’s work for Donald Trump during the 2016 election.
The article claims that Cambridge Analytica harvested data from people’s Facebook profiles.
Here’s the problem though: This is almost impossible. The vast majority of people’s Facebook data is private. This data can only be seen by their friends or if they choose to reveal it to a company.
Of course you could try and scam your way to people’s data. You could create thousands of fake Facebook profiles. Then try to befriend all 128m American voters. Then copy their data. While this is possible on a small scale, Facebook is generally good at weeding out large scale scams like this.
[Update: The March 2018 revelations about Cambridge Analytica revealed that they did managed to get 270,000 people to use an app that harvested around 50m people’s Facebook data. This has been made impossible since April 2015 due to an API change. So while mass scraping of data is now hard – and was in the run up to the 2016 election – it was possible previously. Note that the decay rate of data quality will make a big difference to whether this data was useful in 2016]
No, Facebook doesn’t know more about you than your partner
A related example is almost any article that claims that companies know huge amounts about you. Yes, there is a lot of data, but it’s not always meaningful.
To test this visit Amazon. Then ask yourself how good their recommendations are. They’re probably ok – but not amazing. Amazon have more data to make recommendations about what to buy than virtually anyone else. Yet they struggle to get it right. To complete your personal test visit Google, Facebook and Apple’s websites. Have a look at how well personalised they are.
Why is this?
The vast majority of data is either private, anonymised or just not very useful. If you want a guide to what’s possible then have a look at the world’s best companies at selling data. Simply create an advertising account on Google. Or look at the business sections of Facebook, Amazon, Experian and Axciom.
Impracticality is closely linked to impossibility.
AI is currently going through a bubble that is often based on impracticality. Read reports on autonomous cars and you’d think every taxi driver in the world was about to lose their job. Then look at Uber’s current testing, which recently leaked, and you’ll find that humans had to take over the driving every 0.8 miles on average. That’s a long way from a technology that is going to satisfy regulators and insurance companies.
Illegal is also impossible
A final angle on impossibility is the law. Is the technology legal? And does it comply with relevant social network policies? Data protection laws, and discrimination laws, might simply make it illegal. And if a company scrapes millions of people’s data from Facebook they can expect to be sued.
Law two: Value – New technology isn’t worth the effort
New technology might be practical and legal. But it often isn’t worth the effort compared to easier and cheaper approaches.
There are lots of simple ways to improve digital campaigns, because they are simple to test. It’s common for testing to increase the effectiveness of a campaign by 10-20%, quickly and at low cost. So the second law is that technology may not be a better use of your time than simpler alternatives.
Another way to look at this law is to ask if you have exhausted your current testing programme. Have you got the right strategic choices? Are you using the right channels, for the right purposes? Have you optimised your customer journey, from acquisition to conversion? Is your creative optimised?
Even tech companies don’t always use cutting edge technologies
I recently reviewed the digital marketing programmes of five of the world’s leading marketing automation providers. If anyone is going to be using cutting edge technology, it’s this group. Yet all of them fail on at least one basic measure. Several have very poor quality email programmes, with virtually no personalisation. Some of them don’t re-market to people who visit their website. Their websites are generic even when they know something about you.
Why is this? Some of it is inertia. But it also reflects a judgement that new technology isn’t worth the effort of implementing it.
Apply the second law to claims that people can predict the future (e.g. the X-Factor). You could try through social listening but that’s hard to get right. It’s easier, and probably cheaper, to commission a traditional poll.
Law three: Reach – Technology doesn’t reach enough people to work
A valuable technology has to reach, and influence, enough people to justify its cost. Reach comes down to two things:
Penetration – What proportion of the target audience encounter your technology?
Time use – How long do your audience spend with the technology?
The problem with anything new is that not many people use it. By definition. So there needs to be a route for it to penetrate its market.
This can sometimes happen organically. But most of the time this needs time and marketing.
Apply the third law to VR discussion and something becomes clear. Most VR won’t get to enough people to make a difference. Why? Because there’s no distribution system for it, unless it’s through Google, Apple, Facebook or Amazon.
What is the most hyped technology?
Look at the three laws and what areas of technology look overhyped currently?
Not enough reach, and not much evidence of impact in many cases.
AI that’s implied to be fully automated. If you re-cast these claims as ‘software tools making people smarter’ then there are plenty of sensible cases. But implying that you have an artificial intelligence, when you just have programmatic media buying, you’re pulling a con.
Almost all discussions of Social media that don’t mention reach.
‘Data science’ that claims to know people better than they know themselves.
 There’s also a claim that Cambridge Analytica used ‘machine learning to “spread” through their networks’. This is a confused claim, that might charitably be read to mean that Cambridge Analytica used machine learning to understand what content was most liable to be shared on social media.
Saving 20% of your digital content budget is one of the easiest ways to increase your profit margins.
I had drinks with a couple of industry contacts recently. One works in a large, successful, multi-national. The other, like me, has spent the last decade working in-house and at agencies for big companies.
We started by discussing all the complex, cutting edge industry stuff. Things like Artificial Intelligence.
But we then also agreed that it was pointless for many organisations. Because most of hadn’t fixed the basics first. Savings are readily had almost everywhere.
Almost everyone we’ve ever worked with can get an, almost, free lunch.
Here are three easy ways to fix the basics, and save up to 20% of your digital marketing costs.
Saving on tools
Most organisations now pay for a wide range of tools – with overlapping capabilities. Audit them and you’ll usually find unused or under-used licenses.
Typically these will include tools for:
Storage and servers.
Publishing (website, social).
These are all now easy to buy online. So many bits of your organisation will have bought them. Usually without coordination.
The simplest way to identify what you are buying is to ask your own finance department for a list of all spending of more than, say, £500. Or just ask everyone to report what they are buying.
Most tools are being bought for a reason. But if you look at what they are actually used for, you’ll find that you’ve got the wrong set of tools. Even if the set of tools is exactly right, you can usually secure a significant discount from tool providers every year when you renew your contract.
Savings: 1-5% of your costs.
2. Stop pointless activities
This sounds obvious. But most organisations have significant resources devoted to activities with very low impact.
Organisations often don’t have consistent reporting on what they are achieving. For instance it’s common to see Social channels which get resources, but virtually no reach. Stop this pointless activity and you have savings at almost zero cost.
Here is an analysis of a Twitter account, which took me 10 minutes. This account gets over 4 million impressions per month. Content type A (Pictures, with no caption) gets 3 impressions on average, Content type B (direct replies) gets 1,248 impressions and Content type C (normal tweets) gets 13,354 impressions.
Type A tweets accounted for 4% of the 380 tweets from this account in one month. Type B tweets were another 24%. Eliminating type A and cutting type B to necessary replies (maybe half of the total), would save 16% of resources, but only cut reach by 1%.
Common activities that are pointless include:
Channels which don’t have any impact – either because of mismatch with your audience’s behaviours or low reach.
Types of content which have no impact on your audiences. This is harder to know than it sounds, because a lot of content doing a brand job will have low engagement.
Savings: 10-15% of your costs.
3. Savings from less duplication
Modern organisations constantly create marketing content. Yet staff turnover and organisational siloes mean that once created it’s often lost in practice. These problems are even worse when more than one country is involved.
This means that content constantly gets re-created. In some organisations I’ve seen up to 40% of content being inappropriately created. In a large organisation this can be millions of pounds a year wasted.
Intranets are meant to solve this. But they have almost always failed until recently. New solutions, especially Google Drive and Slack, are now making it easy to share and find content.
This isn’t quite a free lunch. While Slack and Google Drive are cheap, they aren’t always free. And there’s some work to ensure that people adopt them, tag documents appropriately and get good habits.
Savings: 10-15% of costs
There are many other areas you can save on. But to start with, saving on these three areas is a lot easier than using cutting edge technologies.
Beyond Twitter, and the odd private Facebook group, I find that almost a lot of the most useful things I hear come from email lists. So I’ve gone through my favourite lists to find my top five.
Lara O’Reilly’s CMO Today newsletter for the Wall Street Journal has recently become one of my favourite emails, combining the most important business news relevant to marketing with consistently superb puns.
Matt Muir’s weekly emails are an amusing mixture of rants, music and technology. But what makes them really stand out is his list, every week, of major social network changes. The number of Facebook announcements weekly alone is boggling – add in the other social networks and it’s astounding Matt keeps on top of them. But he does – and we all benefit.
Benedict Evans is about as knowledgeable as you get on the mobile industry – his collection of blogposts / podcasts, and links provides critical analysis of what’s going on – based on solid research and thinking, with a good dose of scepticism.
Azeem Azhar’s weekly round up of cutting edge technologies, especially focused on AI, is consistently fascinating. On top of a day job, being an FT columnist, he somehow has time to organise what are meant to be (I’ve never be able to go to one) excellent dinners with leading thinkers.
James is a former colleague and mate. But he’s also be writing this superb mix of geekery for a few years – rightly making him one of the most respected people in the UK Social industry. I have really no interest at all in half of his stuff – that’s about superheroes and stuff like that. But the other half mixes feminism, social media developments and technology rather brilliantly.
“The value of the top 10 corporations was $285tn (£215tn), beating the $280tn worth of the bottom 180 countries, which include Ireland, Indonesia, Israel, Colombia, Greece, South Africa, Iraq and Vietnam.”
Even, usually sensible, commentators like Scott Galloway use this comparison.
It’s a simple enough story. It goes something like this:
Trump’s data agency, Cambridge Analytica, gathered 5,000 data points on everyone. They used this to psychologically profile people, and deliver highly personalised advertising online. This exploited your character, fears and interests. And this swung the election for Trump.
Dig under the skin and this story has a few flaws. Using Cambridge Analytica’s own data, we can see that it probably didn’t swing the election.
To understand why, we need to kill a myth. Which is that Trump’s campaign knew how individuals behave and think in intimate detail.
This requires Trump’s campaign to have abundant data on millions of voters.
Privacy & Cookies Policy
Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information.
Any cookies that may not be particularly necessary for the website to function and is used specifically to collect user personal data via analytics, ads, other embedded contents are termed as non-necessary cookies. It is mandatory to procure user consent prior to running these cookies on your website.